PERFORMANCE EVALUATION OF OPENCV AND GAZETRACKING MODELS FOR AUTOPROCTORING
Saidov M. K. & Rustamov U. A.
Annotatsiya
This paper presents the results of the evaluation of two autoproctoring or unproctoring learning performance model and algorithm combinations. “Model” refers to the use of any user verification or computer monitoring software, and “algorithm” refers to the choice of procedure for dealing with cheating behavior once detected. For months each model/algorithm combination monitored learning performance data during exams taken in two college courses at a community college. During these months the teacher published learning model/algorithm combinations and punished those found violating the learning agreement. Exam scores and the percentage of exams passed fairly both significantly decreased once model/algorithm combinations were published, while the percentage of exams passed fairly and the percentage of exams passed increased in the months after punishments were enforced. However, the percentage of learning performance fairly remained low. These results suggest that the protection provided by learning agreements decreased once learning model/algorithm combinations were published but increased once punishments were enforced [1]. It is anticipated that teachers will have access to the learning model/algorithm combinations published in this evaluation, though the effectiveness of these learning model/algorithm combinations may depend on the context in which they are used. A teacher’s chance of catching a student cheating may range from “high” to “low”, and the deterrence effect outlined in the introduction will be stronger for high-chance teachers. Also, a model’s and student’s chance of catching cheating may range from “high” to “low”, and likewise, the deterrence effect will be different for each combination of model and student chance.
Adabiyotlar
- Y. Feinman, "Security Mechanisms on Web-Based Exams in Introductory Statistics Community College Courses," 2018. [PDF]
- J. Sharma, J. Campbell, P. Ansell, J. Beavers et al., "Towards Hardware-Agnostic Gaze-Trackers," 2020. [PDF]
- T. Effenberger and R. Pelánek, "Impact of Methodological Choices on the Evaluation of Student Models," 2020. ncbi.nlm.nih.gov
- F. F. Kharbat and A. S. Abu Daabes, "E-proctored exams during the COVID-19 pandemic: A close understanding," 2021. ncbi.nlm.nih.gov
- M. Ege and M. Ceyhan, "Talent-Interview: Web-Client Cheating Detection for Online Exams," 2023. [PDF]
- A. Gosiewska, M. Bakala, K. Woznica, M. Zwolinski et al., "EPP: interpretable score of model predictive power," 2019. [PDF]